Anomaly Detection of Object-Centric Event Logs Using Event Knowledge Graph
Omnia Osama Amin, Walid M. Abdelmoez, Mohamed Shaheen · 2024
Process mining describes the technologies that use data to help businesses better understand their processes based on captured event logs extracted from information systems, relational databases, or business management software. To apply process mining techniques, event logs are used as input. Traditional process mining techniques associate each event with exactly one object, leading to convergence and divergence problems. Object-centric process mining (OCPM) takes a more extensive approach to process mining by considering many object types and events that involve multiple objects, increasing the data complexity and making it harder to detect anomalies. The event knowledge graph is powerful in visualizing complex relationships and interactions among entities in a graph-based structure. It helps organize data related to events where objects play a central role. It enables a better understanding of connections and patterns within object-centric event logs (OCELs). Detecting anomalies in an event knowledge graph involves identifying unusual patterns, behaviors, or relationships that deviate from expected norms. This paper proposes an approach for integrating OCEL into an event knowledge graph and utilizes graph-based techniques for anomaly detection. Our approach was evaluated on a synthetic event log in the logistics domain.